Re: how to troubleshoot performance problems that trigger kidle-inject

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, Aug 17, 2018 at 1:09 AM, Samuel Sieb <samuel@xxxxxxxx> wrote:
> On 08/16/2018 01:31 PM, Chris Murphy wrote:
>>
>> kernel 4.17, and there's only light fan or no fan usage when those
>> processes spin up.
>
>
> After a little research, it appears that this is related to the
> intel_powerclamp module which is supposed to protect your processor from
> overheating.  What is the CPU temperature?

https://drive.google.com/open?id=1tSbrVqFRz3oj3hgmRkfKJsIUuNViKJsQ

That looks like around 50C if pch_skylake is the actual CPU. The
spikes above the ~54C dashed blue line are the initial loading of web
sites, which doesn't last long and then settles but due in large part
to increasing values for kidle_inject. I don't know the relevance of
that dashed blueline.

>I know you said the fan is low,
> but maybe the fan control isn't working properly?

I think that may be correct, I'm not hearing much fan at all during
these events. And this is relatively recent behavior in the past 6
months. When I first got the laptop and was running Fedora 26/27 I
didn't have this problem, fans came right up, just like on Windows 10.
Windows 10 might have even more aggressive fan behavior.

>  Or it might be the GPU
> temperature.  You could try "rmmod intel_powerclamp", but be careful.  Make
> sure you check your temperatures.

I'm leary of changing that until I can figure out why the fans aren't
coming up. Or maybe the kidle_inject is preventing fans from coming on
in the first place, the documentation suggests that's one of the use
cases for powerclamp.

>
> More info:
> https://www.kernel.org/doc/Documentation/thermal/intel_powerclamp.txt

"cur_state allows user to set the desired idle percentage. Writing 0
to cur_state will stop idle injection."

[root@f28h cooling_device9]# echo 0 > cur_state
[root@f28h cooling_device9]# grep . *
cur_state:33
max_state:50
grep: power: Is a directory
grep: stats: Is a directory
grep: subsystem: Is a directory
type:intel_powerclamp
[root@f28h cooling_device9]# grep . *
cur_state:29
max_state:50

Sometimes writing 0 immediately stops the inject. But then it returns.
So this isn't working as I'd expect. Anyway, even when there is no
inject (screenshot)
https://drive.google.com/open?id=1drbHJRfCyiPmNKXP_vO4mqyqTKKGplUL

Fans only very slightly spin up at above 60C, the two bumps at ~65/70 samples.

This is very different on Windows 10 where even a single web page
loading kicks up the fans into medium high, and they can go even
higher, and I get no sluggish behavior at all. So something is really
not tuned correctly, but also the problem is made worse on Fedora with
a 2nd display, and it may also be GPU related as in the GPU offloading
isn't enough with browsers on Fedora (?) whereas the GPU does more
efficient work on Windows 10. It doesn't appear to be any one
particular thing but the lack of cooling with fans is sorta top on the
list of suspicious.



-- 
Chris Murphy
_______________________________________________
devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx/message/4EBNUHE6RQ7L6SKKG2CVUEEBEO3HQD6G/




[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [Fedora Announce]     [Fedora Users]     [Fedora Kernel]     [Fedora Testing]     [Fedora Formulas]     [Fedora PHP Devel]     [Kernel Development]     [Fedora Legacy]     [Fedora Maintainers]     [Fedora Desktop]     [PAM]     [Red Hat Development]     [Gimp]     [Yosemite News]

  Powered by Linux